The effects of ketamine on symptoms of depression and anxiety in real-world care settings: a retrospective controlled analysis

Ketamine intravenous therapy (KIT) has been shown in controlled trials to produce rapid antidepressant effects when delivered as a limited series of infusions, and prior studies also report benefit for anxiety symptoms in treatment-resistant depression and anxious bipolar depression. Real-world clinic populations, however, are typically more heterogeneous than trial samples, often presenting with comorbid anxiety and depression, and many clinics now use KIT protocols that diverge from those tested in controlled research. As a result, there is limited controlled comparison of mood and anxiety outcomes, and of the stability of those outcomes, from KIT delivered in routine care settings. Hietamies and colleagues set out to measure the impact of KIT on self-reported depressive and anxiety symptoms across both induction and maintenance phases in a large, multisite, real-world dataset. The study additionally compared KIT outcomes to two external datasets — one observational sample of patients evaluated but not treated with ketamine and one clinical trial sample initiating standard antidepressant monotherapy — to contextualise treatment-specific effects and address confounds such as regression to the mean and expectancy effects. The investigators hypothesised that KIT would show superior symptom reductions compared with these control samples and that benefits would remain stable during maintenance treatment.

This was a retrospective, de-identified, HIPAA-compliant pooled dataset of patients treated with KIT at ten Actify Neurotherapies clinics across nine US states between August 2017 and March 2020. The initial extraction comprised 2,758 patients who received at least one KIT-related visit; patients were largely self-referred and paid out-of-pocket. Available demographic information included age, sex, and clinical diagnoses, but the extracted text does not clearly report details such as concurrent psychotropic medications, psychosocial support, or ketamine dosing parameters. Symptoms were measured using patient self-report: the Quick Inventory of Depressive Symptomatology–Self Report 16-item (QIDS; range 0–27) and the Generalized Anxiety Disorder 7-item scale (GAD-7; range 0–21). Questionnaires were completed at intake, at the first maintenance visit (typically the seventh infusion), and at subsequent visits. The investigators defined induction a priori as at least four infusions within 28 days, although due to the measurement schedule all included subjects in the analysed outcomes received at least seven infusions; any further infusions were classified as maintenance. Inclusion for the primary clinical outcomes required a baseline QIDS and GAD-7, completion of the induction series, and a follow-up QIDS and GAD-7 at the start of maintenance. The study team notes that the extraction does not clearly state ketamine dose/concentration or the presence/absence of concurrent psychological therapy. For comparison, two external datasets were used: an observational ‘‘no ketamine treatment’’ sample of 276 subjects who had a KIT intake but received zero KIT visits and had PHQ-9 outcomes at baseline and 21–42 days; and data from 1,008 participants in the iSPOT-D study who were randomised to escitalopram, sertraline, or venlafaxine-XR and assessed with QIDS up to eight weeks. Statistical methods included within-sample and between-group t-tests, chi-square tests for categorical outcomes, Pearson correlations, Cohen's d effect sizes, and multilevel mixed-effects models (random intercepts) with a fixed linear time effect to estimate maintenance-phase trends. To compare samples using different outcome measures, the investigators computed normalized change scores ((follow-up − baseline) / SD baseline). An alpha of 0.05 was used for hypothesis testing and analyses were performed in Python 3.8.

From the initial 2,758-subject dataset, 836 met inclusion criteria for maintenance-phase analyses and 714 (a subset) met criteria for post-induction outcomes (first post-induction visit within 28 days). Attrition during induction was substantial: 533 subjects (19%) had zero induction visits logged, and retention across induction visits fell to 61.3% by visit 6. Subjects included in the post-induction analysis had higher baseline depressive severity (median QIDS 18.0) than those excluded (median 15.7). Across the analysed completer sample, mean total infusions were 13.3 (SD 7.9, median 11). Mean QIDS fell from 18.0 at baseline (SD 4.6) to 9.6 at post-induction (SD 5.3), a mean change of −8.36 (SD 5.4), which was statistically significant (t = 41.6, p < 0.001, 95% CI [−8.76, −7.97]) with a large effect size (d = −1.56). By the QIDS, n = 351 (49.2%) met response criteria post-induction and n = 187 (26.2%) remitted. For anxiety, mean GAD-7 declined from 15.2 (SD 5.7) to 8.6 (SD 5.6) post-induction, a mean change of −6.68 (SD 5.7), also statistically significant (t = 31.0, p < 0.001, 95% CI [−7.09, −6.26]) with d = −1.17; response and remission rates were n = 337 (47.5%) and n = 183 (25.6%), respectively. Baseline severity correlated negatively with change (QIDS r = −0.45; GAD-7 r = −0.51), indicating greater absolute improvement among those with higher baseline scores. Change scores for QIDS and GAD-7 were strongly correlated (r = 0.60), and 90–95% of participants were classified as the same or improved on each measure. Suicidal ideation (QIDS item 12) was present in 597/731 (81.6%) at baseline; among these, 409/597 (70%) reported improvement in SI at post-induction and 290/597 (49%) reported no SI. Across all patients, 41 (5.7%) showed worsening of SI at follow-up, mostly by a single-point increase. When compared with the observational non-ketamine sample (n = 276 PHQ-9 assessed), that control group showed a reduction in PHQ-9 (b = −5.34, d = −1.00). The KIT induction sample experienced a significantly larger normalised reduction in depressive symptoms (t = −10.01, p < 0.001; between-group d = −1.03). Worsening of depression at follow-up occurred in 4.5% of KIT patients versus 16.7% in the non-treatment sample (chi-square, p < 0.001). Compared with the iSPOT-D antidepressant-treated cohort (n = 1,008), which showed a QIDS reduction of b = −3.48 (d = −0.92) at eight weeks, the KIT group again had a greater normalised reduction (t = −16.26, p < 0.001) and fewer individuals with worsened depression (KIT 4.5% vs iSPOT-D 9.5%, p < 0.001). Effect-size differences favouring KIT were largest at earlier time points (d = −1.09 at 2 weeks) and remained present though attenuated by eight weeks (d = −0.62). During maintenance (visits 7–26), both QIDS and GAD-7 scores showed small but statistically significant increasing linear trends per visit (QIDS b = 0.063 per visit, p < 0.001; GAD-7 b = 0.066 per visit, p < 0.001). These per-visit changes were small relative to score variability, with per-visit Cohen's d estimates for QIDS ranging roughly between −0.36 and 0.16, and for GAD-7 between 0 and 0.25. Response and remission rates were relatively stable through visit 16 (when sample retention declined below 20%), with response rates for QIDS between 44.9% and 48.0% and for GAD-7 between 45.8% and 50.1% across visits 7–16. Among induction non-responders who continued into maintenance (n = 386 QIDS, n = 400 GAD-7), cumulative delayed-response rates reached 38.3% for QIDS and 40.5% for GAD-7, with the majority of delayed responders achieving response by visit 12. Overall across post-induction visits, clinical response was achieved in approximately two-thirds of subjects (QIDS 68.3%, GAD-7 68.7%).

Hietamies and colleagues interpret these results as evidence that KIT, when delivered as a series of infusions in routine care, is associated with robust reductions in self-reported depressive and anxiety symptoms and with reductions in suicidal ideation among those who complete induction. They emphasise that symptom improvements observed post-induction were clinically meaningful and that many patients who did not respond during induction achieved delayed response during early maintenance, leading to cumulative response rates near 67%. The investigators acknowledge important limitations of retrospective real-world data. High attrition during induction (~40% dropped out prior to a minimum of four infusions) likely inflates observed effect sizes among completers; the dataset was also limited by incomplete patient-level information and inability to fully account for confounds such as concurrent treatments, non-specific expectancy effects, and regression to the mean. To address some of these concerns, the authors compared KIT outcomes to two external datasets and found consistently larger improvements and lower rates of symptomatic worsening in KIT patients than in both an observational non-treatment sample and an antidepressant-treated clinical trial cohort, noting that these comparisons are imperfect but provide context for non-specific influences. Regarding comorbidity, the study found no evidence that high baseline anxiety reduced KIT's antidepressant effect; reductions in anxiety and depression were concordant. The authors situate their findings within prior KIT real-world and controlled studies, noting similar response and remission rates and emphasising reproducibility across clinical populations. Finally, they caution that maintenance-phase inferences are limited by attrition at later visits and that the retrospective design and missing details (for example, dosing and adjunctive psychosocial support are not clearly reported in the extracted text) constrain causal interpretation. Despite these caveats, the authors conclude that their real-world evidence supports KIT as an effective treatment option for depressive and anxiety symptoms when administered as a series of infusions.